Ollama shipped v0.35.0 on 2026-09-28 (GitHub prerelease): Decision Models via TypeSafe’s Jev API at /v1/systemone. Instead of free text, models answer choice/noul/score questions with options, probabilities, and confidence—for ticket triage, model routing, and classification. First models: Nimble (Bespoke Labs) and Tev1 (Together AI). ~7 commits / 66 files vs v0.34.4; Settings no longer blocks on discovery; macOS update menu and stalled MLX downloads fixed.

Key Takeaways

  • ✓v0.35.0 (prerelease) adds Decision Models at POST /v1/systemone (TypeSafe Jev)
  • ✓Question types: choice (with probabilities), noul, score — not free-text generation
  • ✓First models: Nimble (Bespoke Labs) and Tev1 (Together AI) on ollama.com/library
  • ✓~7 commits / 66 files vs v0.34.4; Settings/macOS update/MLX download fixes
🧭

Finished reading? Explore benchmark rankings & pricing

Real-world SWE-bench scores & $20/mo vs API cost break-even calculator

🏆 Leaderboard→💰 Compare Plans↗
🔗

Project Links & Resources

Direct Access
Direct access to official project resources and documentation
🐙
GitHub Release v0.35.0
github.com/ollama/ollama/releases/tag/v0.35.0
↗
🐙
Compare v0.34.4...v0.35.0
github.com/ollama/ollama/compare/v0.34.4...v0.35.0
↗
🚀
Ollama library: Nimble
ollama.com/library/nimble
↗
🚀
Ollama library: Tev1
ollama.com/library/tev1
↗
📝
TypeSafe (Jev)
typesafe.ai/
↗
🐙
ollama/ollama repository
github.com/ollama/ollama
↗
🌐
ollama.com
ollama.com/
↗
🔬

In-Depth Technical Analysis

Core Background & Industry Pain Points

Agents often force triage/routing/classification through chat completions, then regex the prose—expensive and brittle. Decision workloads need structured choices with probabilities/confidence. TypeSafe’s Jev-style APIs exist for this; Ollama’s main path was still generative chat.

Architecture Highlights & Internals

v0.35.0 adds Decision Models at POST /v1/systemone with model, state, and questions. Types: choice, noul, score. First library models: Nimble and Tev1. Example: checkout-500 ticket → bug ≈0.978 probability, confidence ≈0.89, plus usage tokens. Also: Settings opens without waiting for discovery; macOS update menu/icon fix; stalled MLX downloads fixed; deprecated typical_p warns instead of failing.

Authoritative Benchmarks & Measured Scores

No public SWE-bench/MMLU table in the release notes. Signal is ~7 commits / 66 files vs v0.34.4 and the example probability fields. Calibrate on your own labels; tag is a GitHub prerelease.

Developer Hands-on Guide

Install v0.35.0, ollama pull nimble, call /v1/systemone, wire choice/probabilities/confidence into routers. See release, Nimble, Tev1, TypeSafe.